Autonomous Engineering Applied to Investment Casting Process. ICI Conference October 15-18, 2017

Transcription

1 Autonomous Engineering Applied to Investment Casting Process ICI Conference October 15-18, 2017

2 Overview What is Autonomous Engineering? Traditional simulations vs new approach Case Study #1 Using Autonomous Engineering to evaluate two pattern layouts Case Study #2 Adjusting Thermal Property datasets for Investment Casting Shells Case Study #3 Evaluating a gating approach using Autonomous DoE

3 From Simulation to Autonomous Engineering

4 Investigating Solidification MetalTek International (US) Casting geometry with premanufactured cone (Shell hidden) No porosities detected solidification pattern Indicating feeding paths Shell surface temperatures after ~ 21min (end of solidification)

5 Investigating Stresses and Cracks Resulting Cold Crack Risk Temperature Distribution Local tensile strength

6 Investigating Stresses and Cracks Cold Crack Result

7 Reoxidation inclusions in steel alloys Inlet

8 Robust Casting Designs and Processes a Black Box?... everything happens at the same time flow cooling heat flow gas segregation Many influencing variables can only be indirectly measured and can not be controlled convection composition processing stresses solidification metallurgy distortion heat treatment

9 Why autonomous engineering? Improve the solution Quality criterion / Objective 1 optimal solution Never touch a running system vs. There is always a better solution! initial concept manually optimized specification Quality criterion / Objective 2

10 Why autonomous engineering? Support continuous improvement processes SOP: large process variation running production: process variation reduced (process robustness) running production: robust process + optimized operating point (quality, costs, productivity)

11 Case Study #1 Autonomous Engineering applied to two pattern layouts

12 Wax model and simulation setup 3D Simulation Model with stand and ground

13 Runner/Feeder Layouts A and B A B

14 Process Parameters Cast Material: High-Alloy CrNi-Steel Pouring Temperature: 1600 C Pouring Time: ~ 5s Ceramic Shell preheated at 850 C Cast in ambient environment Down sprue isolated through topping

15 Filling Simulation Results

16 Metal flow and temperatures during filling

17 Re-oxidation inclusions Svoboda, J.M., et al., Appearance and Composition of Oxide Macroinclusions in Steel Castings, AFS Transactions vol. 95, 1987, pp The majority of inclusions in steel castings result from reoxidation and turbulence during mold filling

18 Other Filling Results Air entrapped by the metal front Massless particles showing flow patterns

19 Solidification Simulation Results

20 Radiative Heat Exchange A B

21 Temperatures during Solidification Metal Temperatures Color Scale: Shell Temperature Color Scale: B

22 Solidification Path of Layout A A Fraction Liquid with areas which no longer can be fed are shown invisible

23 Critical Areas during Solidification Layout A A Regions isolated from feed paths Areas with macroscopic shrinkage

24 Casting defects Actual Defects and Feeding Paths Layout A A Insufficient feed paths

25 Critical Areas during Solidification Layout B B Spots Regions withisolated from Solidification feed paths Areas with macroscopicporosity shrinkage

26 Summary Runner/Feeder Layout A does not provide feeding paths which result in a sound part Layout B shows improvements concerning tendency to form porosity, but some indications remain To get a casting whose quality is robust in spite of process fluctuations, and to better understand the influence of geometrical changes on feeding, it is decided to perform a set of virtual experiments with different parameter variations Casting Layout B is the subject of the further investigation

27 Autonomous Engineering through virtual Design of Experiments for Layout B

28 Parameter Variations in Layout B DoE 1: Influence of changing process parameters on solidification? DoE 2: Support feeding by changing dimensions of runner and gates? : Delayed pouring: Wait time: 10s, 65s, 120s Shell preheat temperature: 800 C, 900 C, 1000 C Pouring temperature: 1500 C 1650 C, step 50 C Geometry as before Process parameters fixed.

29 Design of Experiments 1, Changing Process Variables Goal is to examine influencing process parameters on the critical areas during solidification Calculation covers all 36 possible combinations Comprehensive comparison and utilization of all simulation results Quantitative assessment in charts and tables

30 DoE1: Effect of Pouring Delay on shell temperature Example: Fixed Temperatures for Melt (1600 C) and Shell (900 C) 10s 65s 120s

31 DoE 1: Assessment of Virtual Experiments Ranking according to chosen result values Experiments with higher Melt Temperatures (red, pink) show less tendency for cutting off feed paths Experiments using higher shell preheating (greenish) show both good and bad tendencies. No obvious effect.

32 Tendency for cutting off feed paths DoE 1 Assessment using Main Effects diagram T init Melt Delay time T init Shell In contrast to the melt temperature, the delay time and the shell preheat temperature show no significancy in terms of the chosen quality criterion

33 DoE1: Selected Results of virtual Experiments Areas isolated from feed paths

34 Design of Experiments 2: Geometry Variation Objective is to improve feeding paths into the casting by the following variables: Widening gates Varying runner section Changing down sprue position

35 Tendency for isolated solidification DoE 2, Assessment using Main Effects Diagrams Gate Width Pos. Downsprue Runner Section

36 Tendency for cutting off feed paths Area1 DoE 2, Assessment using Scatter charts Poor Bubble Size: Runner Section Poor Bubble Size: Gate Width It is more important to have a wider gate than having bigger runner sections Good Tendency for cutting off feed paths Area2 Good Tendency for cutting off feed paths Area2

37 DoE 2: Best Compromise Widening the gates prevents the separation of the feeding paths The runner section can be reduced -> better yield Down sprue position has only minor influence

38 Summary of DoE 1 and DoE 2 Virtual Experimentation delivered insight and revealed influencing process parameters For the considered Investment Casting and with respect to the examined criteria (Cutting off of Feed Paths -> Porosity), the process parameter which is most important is the Melt Temperature. The variation of the Delay Time as well as the Shell Temperature is almost negligible The results of geometrical variation shows that as long as the gates are widened, the runner section and the down sprue position have almost no impact

39 Case Study #2 Adjusting Thermal Property Datasets of Investment Casting Shells

40 Introduction Using Simulation for Investment castings. 1. New Jobs where gating needs to be cut into the die before the die is shipped. 2. Prototype jobs where the gating needs to be printed on the part. 3. Old problem jobs to reduce the amount of post processing such as welding. How can we increase the accuracy of our simulations? X-ray results vs. Simulation results. Study input variables. No general dataset can be used to describe every foundries Ceramic Shell. shell compositions particle size distribution processing parameters Dip # Slurry Stucco 1 Primary Zircon/Silica Zircon 2 Intermediate Silica 50/100 3 Back up silica 30/50 4 Back up silica 30/50 5 Back up silica 30/50 6 Back up silica 10/30. 7 Back up silica 10/30.

41 Experiment introduction Matching Cooling Curves In the Foundry: We will measure the temperature inside the mold cavity and outside the shell to see how heat passes through the shell. In Simulation: We will use inverse optimization to incrementally change thermal conductivity and specific heat of our shell until our temperature vs. time curves match that produced in the foundry.

42 Design of Experiment Thermocouple setup Shells. done by MS&T. Materials: Steps: K, B, and S type thermocouple wire Quartz Glass Tube Alumina Double Bore tube Mini connectors Mold in a Quartz glass tube Build Shell Autoclave Drill and Place K type wire just under shells outside surface Attach insulation to the outside of the shell Fire shell and observe shell temperature Remove shell from oven Insert thermocouple probe into glass tube Pour the shell and record the data.

43 Molding in Glass Tube

44 Thermocouple Probe Materials S and B type wire S and B type extension wire Quartz glass tube Alumina double bore tube Connectors based on your data logger

45 Inserting K type wire and Insulation

46 Collecting Data

47 Inverse Optimization: Setting Design Variables

48 Inverse Optimization: Setting Design Variables

49 Assessment: Curve Comparison

50 Assessment: Main Effects Plot Based on on Gradient Integral objective Thermal Conductivity Specific Heat Capacity

51 Assessment: Curve Comparison

52 Simulation Results Temperature F Thermal Conductivity (W/mK) Temperature C Best Design Cooling Curve Current Data set Current Dataset cooling curve Seconds

53 Summary The Main Effects Plot quickly showed that the thermal conductivity and the specific heat capacity had the biggest impacts on matching the simulated cooling curves with the measured cooling curves Fine tuning those properties provided simulated curves to match measured data The porosity prediction more closely matched what was seen in production once the proper thermophysical properties we determined using inverse optimization This provided more confidence in the simulated results and reduced sampling time

54 Case Study #3 Evaluating a gating approach using Autonomous DoE

55 Gating System/Casting Layout The goal is to achieve smooth filling that will produce good surface finish and eliminate repairs

56 Process Parameters Defined variables that can be changed Cast Material: CuSn12 Tin bronze Pouring Temperature: 2000 F Pouring Time: ~ 2s Ceramic Shell preheated at 800 F

57 Filling Simulation Results

58 Metal Temperatures During Filling the objective is to track metal front temperatures to avoid misruns or cold laps Liquidus Initial Pour Temp Solidus Liquidus

59 Metal Front Velocities During Filling the objective is to track metal velocities to reduce turbulence and inclusions

60 Other Filling Results Air entrapped by the metal front Weightless particles showing flow patterns

61 Solidification Simulation Results

62 Solidification Path Fraction Solid Result Transparent areas are no longer open to feeding

63 Locations of Significant Themal Centers Regions isolated from feeding through the gating system Areas with macroscopic shrinkage porosity

64 Virtual Experimentation through Design of Experiments

65 Parameter Variation Design of Experiment setup parameters Possible optimization paths: DoE 1: Influence of changing process parameters on solidification? Pouring speed: Ladel height: 3, 6, 9 DoE 2: Gating modifications?: Shell preheat temperature: 500 F, 600 F, 800 F Pouring temperature: 1800 F 2250 F, step 20 F Baseline geometry Process parameters fixed Goal is to examine the influence of process parameters on the critical areas during filling and solidification

66 Parralell Coordinates Each line represents a design that has been run Pouring Speed Pouring Temperature Shell Temperature Minimum Metal Temperature Porosity

67 Parralell Coordinates We want to look at designs whos minumum temperature exeeds the liquidus Pouring Speed Pouring Temperature Shell Temperature Minimum Metal Temperature Porosity

68 Parralell Coordinates Which designs have the lowest porosity Pouring Speed Pouring Temperature Shell Temperature Minimum Metal Temperature Porosity

69 Scatter Plot Porosity vs Minimum Metal Temperature Porosity Minimum Temperature

70 Correlation Matrix Shows how variables impact the objective Objectives Variables

71 Summary The DoE showed that a slower pouring speed and a higher pouring temperature was needed to produce a part with the lowest porosity Porosity will likely increase if multiple parts are poured in the same heat The shell temperature also had an impact on the porosity, but the impact was less than the pouring speed and pouring temperature The information gained from running an Autonomous DoE can assist in establishing a process window to ensure consistent quality from the process

72 Conclusion Data driven decision making capabilities available using Autonomous Engineering allows the engineer to increase the quality and consistency of their products while having the ability to focus on reducing cost through inventive design considerations Using the capability to autonomously run virtual designs of experiments optimizations to find optimal casting process parameters and design feature combinations to make quality castings at minimal costs is the core benefit of autonomous engineering methodology integrated into the casting process simulation software

73 Thank you